Simple script to grab tweets with polling info and process them to see what leads to retweets.

polling_tweets.R

library(rtweet)
library(data.table)
library(ggplot2)
library(stringr)
tweets <- as.data.table(get_timeline("PpollingNumbers", n=3200))

tweets <- tweets[is_retweet!=TRUE,]
tweets[,nat:=grepl("National", text, ignore.case = T) &
         grepl("Biden", text, ignore.case = T) & 
         !grepl("students", text, ignore.case=T) &
         !grepl("favor", text, ignore.case=T) &
         !grepl("matchup", text, ignore.case=T)&
         !grepl("non-white", text, ignore.case=T)&
         !grepl("african", text, ignore.case=T)&
         !grepl("general|genral", text, ignore.case=T)&
         !grepl(" GE", text, ignore.case=F)&
         !grepl("hispanic", text, ignore.case=T)&
         !grepl("head-2-|head-to", text, ignore.case=T)&
         !grepl("consider", text, ignore.case=T)&
         !grepl("trust", text, ignore.case=T)&
         !grepl("job", text, ignore.case=T)&
         !grepl("won", text, ignore.case=T)]
tweets[,biden:=as.numeric(str_match(str_match(text, "Biden[: ]*\\d*"), "\\d+"))]
tweets[,sanders:=as.numeric(str_match(str_match(text, "Sanders[: ]*\\d*"), "\\d+"))]
tweets[,comp:=cut(sanders-biden, breaks=c(20,-1,-5,-10,-15,-100), 
                  labels=rev(c("Sanders Tied or\nBeating Biden", 
                           "Biden +1 to +5", 
                           "Biden +6 to +10", 
                           "Biden +11 to +15", 
                           "Biden +16 or more")), ordered_result = T)]
tweets[,raw_dif:=sanders-biden]
# View(tweets[nat==TRUE,.(date,retweet_count, sanders, biden, comp, text)])
tweets[,date:=as.Date(created_at)]
library(ggthemes)
# tweets$comp
ggplot(tweets[nat==TRUE & !is.na(comp),], aes(x=date, y=retweet_count, fill=comp)) + 
  geom_point(pch=21, color='gray') + 
  theme_minimal() + scale_y_log10() +  geom_smooth(aes(col=comp),
                                                   se = F, 
                                                   formula=y~s(x), 
                                                   method='gam') + 
  scale_color_brewer("Results in Tweet", type='seq', palette = 'YlGnBu',
                    guide = guide_legend(reverse = TRUE)) + 
  scale_fill_brewer("Results in Tweet", type='seq', palette = 'YlGnBu',
                    guide = guide_legend(reverse = TRUE)) + 
  labs(y='Number of Retweets', caption='Data collected from @PpollingNumbers', 
       x='') 

ggsave('polling_rts.png', height=4, width=7, units='in')


ggplot(tweets[nat==TRUE & !is.na(comp),], aes(x=raw_dif, y=retweet_count)) + 
  geom_point() + 
  theme_minimal() + scale_y_log10() +  
  geom_smooth(se = F, 
              formula=y~s(x), 
              method='gam')  +
  labs(y='Number of Retweets', x='Sanders-Biden', 
       caption='Data collected from @PpollingNumbers')